The authors wanted to test the long-term effectiveness of immunoisolation in the NOD mouse, an adequate model of Type I diabetes. Recipients were diabetic female NOD mice with sustained plasma glucose levels above 400 mg/100 ml. They were grafted intraperitoneally with permselective hollow fibers seeded with human insulinoma (n = 6), rodent insulinoma (n = 6), or human islets (n = 6). Controls were 15 nontreated diabetic females and 10 nondiabetic male NOD mice implanted with nonseeded macrocapsules. Electron microscopy (rectangular crystalline nucleoids characteristic of B-cells), in vitro release of insulin (during abrupt changes in glucose concentration from 40 to 400 mg/dl and return, M +/- SD insulin levels were 122 +/- 5 uu/ml and 315 +/- 17 at low and high glucose), evidence of binding hormone receptors (VIP and GIP, the binding sites being coupled with adenylyl cyclase stimulation) were used to assess the quality of the transplant. Survival was always prolonged in grafted animals. Taking complete and long-term (less than 3 months) correction of hyperglycemia as the criterion, the success rate was about 50% as observed in streptozotocin rats. Splenocytes isolated from cured and not cured recipients and injected into irradiated nondiabetic male NOD mice were always able to transfer the disease. Our bioartificial pancreas is efficient in preventing the recurrence of the disease in autoimmune diabetes.